models/neurones_flash.py

"""
Neurones Flash 2.0
==================
NeuraPrompt's fastest model. Powered by openai/gpt-oss-120b.
Best for: quick answers, chat, translation, search.
NOT for: deep reasoning, image analysis.

To upgrade: change `groq_model` and bump `version`. That's it.
"""

MODEL = {
    # ── Identity ────────────────────────────────────────────────
    "id":           "neurones-flash-2.0",
    "display_name": "Neurones Flash 2.0",
    "version":      "2.0",
    "release_date": "2026-03-29",
    "tagline":      "NeuraPrompt's fastest model. Zero wait, instant answers.",

    # ── Speed ───────────────────────────────────────────────────
    "speed":        "ultra-fast",
    "speed_label":  "⚡ Ultra Fast",

    # ── Backend ─────────────────────────────────────────────────
    # Change this line to upgrade to a newer/faster Groq model:
    "groq_model":   "openai/gpt-oss-120b",
    "max_tokens":   2048,
    "temperature":  0.7,

    # ── Capabilities ────────────────────────────────────────────
    # main.py reads these flags to enable/disable features per model
    "can_stream":         True,
    "can_reason":         False,   # No deep think — speed over depth
    "can_vision":         False,   # No image input
    "can_generate_image": False,
    "can_search":         True,
    "can_code":           True,
    "can_translate":      True,
    "can_summarise":      True,
    "is_local":           False,   # Never use local TF-IDF fallback

    # ── Limits ──────────────────────────────────────────────────
    "context_window":  8192,
    "rate_limit_rpm":  20,

    # ── System Prompt ───────────────────────────────────────────
    # This becomes the model's personality in all conversations
    "system_prompt": (
        "You are Neurones Flash 2.0, NeuraPrompt's fastest AI model, "
        "created by Andile Mtolo (Toxic Dee Modder). "
        "Your greatest strength is speed. Give crisp, direct, immediate answers. "
        "Avoid lengthy preamble — get to the point in the first sentence. "
        "Use markdown for code and lists only when genuinely needed."
        "You are limited to only 100-200 words, nothing more, ask the user if they want to ask more questions."
        "Do not format into table, Markdown if you are not using tool search"
        "Ask for clarity if not undertood."
    ),

    # ── UI Hints (frontend reads these) ─────────────────────────
    "badge_color":  "#00e5ff",
    "icon":         "⚡",
    "recommended_for": [
        "quick questions", "live chat", "translation",
        "search", "short summaries", "fast lookups",
    ],
    "not_recommended_for": [
        "deep analysis", "complex math", "long documents", "image tasks",
    ],
}